Cross-modal retrieval aims to bridge the semantic gap between heterogeneous modalities — such as images and text — by learning a shared embedding space for semantically aligned representation. While recent models have achieved impressive performance using large-scale contrastive pretraining and multimodal transformers, several fundamental challenges remain unresolved. These include the lack of interpretable latent alignment, vulnerability to distribution shifts, and instability in semantic correspondence across tasks and domains. In this paper, we propose a novel contrastive representation learning framework designed to enhance both the robustness and interpretability of cross-modal retrieval. Our method incorporates a hierarchical dual-stream encoder that preserves modality-specific structures while enabling semantic interaction through a concept-aligned projection layer. The model is optimized via a contrastive loss with semantic-aware calibration, encouraging consistent feature correspondence across modalities. We provide a rigorous theoretical analysis of the latent projection space, and demonstrate through extensive experiments on MS-COCO, Flickr30K, and RSICD that our approach outperforms strong baselines not only in retrieval accuracy but also in robustness under noise and interpretability via semantic stability selection. The proposed framework is further validated through ablation studies that isolate the contributions of architectural components and training strategies. Our results confirm that semantic disentanglement and hierarchical encoding jointly improve retrieval quality, cross-domain generalization, and feature transparency. The framework offers a scalable and theoretically grounded solution for reliable and explainable multimodal retrieval.

Battery-free Wireless Sensor Networks (BF-WSNs) have emerged as an essential part of Internet of Things (IoT) systems. Although two-dimensional (2D) BF-WSNs have been researched, three-dimensional (3D) ones are more indicative of real-world applications. Achieving$k$-coverage in this scenario is a greater challenge and has yet to be investigated. This paper presents an optimization problem in heterogeneous 3D BF-WSNs to maximize$k$-coverage quality while considering the recharging and sampling rates. We prove the problem is NP-Hard and decouple it into two subproblems. The first optimizes$k$-coverage quality for each time slot without energy constraints. The second ensures that nodes scheduled for operation in each time slot can be immediately replenished with sufficient energy. We prove the near-optimal and optimal solutions to the two subproblems composes the near-optimal solution to the original problem. Following the development of Distributed Iterative Grouping (DIG) algorithm and Adaptive Sampling (AS) method to address two subproblems each, we propose a sleep scheduling algorithm to integrate them and solve the original problem. Simulation results verify the effectiveness and efficiency of the proposed algorithm.

Backscattering enables battery-less and perpetually operating Internet-of-Things (IoT) devices to sweep through ambient electromagnetic waves (in terrestrial networks) or acoustic waves (in underwater networks) rather than generate new waves. The emerging underwater acoustic backscatter communication (UABC) has the potential to avoid the need for underwater nodes to actively emit acoustic signals, thus, reducing their energy consumption. However, there is a lack of a clear design, comprehensive modelling or analysis of UABC systems. This paper studies a UABC system where a high-power transceiver transmits acoustic signals and a piezoelectric UABC node modulates and reflects the incident acoustic signals back to the transceiver. Specifically, a detailed new design for UABC nodes is proposed and a circuit model for the nodes is developed to analyze their acoustic impedance under different materials and operating frequencies. Based on the circuit model, a node load design is introduced to maximize the acoustic reflection coefficients; and a comprehensive communication analytical framework is established and is used to analyze how different UABC parameters, such as piezoelectric material properties, operating frequencies of UABC nodes, transmission distances, ambient noise intensities, carrier wave frequencies, and wind speeds, affect the bit error rate of UABC. Simulation results validate the effectiveness of the proposed designs and models as well as identify the key parameters influencing the UABC performance.

Intelligent omni-surfaces (IOSs) have been considered for assisting outdoor-to-indoor millimeter-wave (mmWave) communications. Nevertheless, the existing works have not adequately investigated how the number or the deployment locations of IOSs should be optimized for serving multiple indoor users. In this paper, we study IOS-assisted outdoor-to-indoor mmWave communications where IOSs are installed in an exterior wall of a building to refract mmWave signals from an outdoor base station (BS) to indoor users that locate among indoor blockages. Given a fixed total number of refracting elements, we formulate an optimization problem to maximize the downlink energy efficiency of the outdoor BS while satisfying the dowlink data rate requirements of the indoor users by jointly optimizing the number, locations and phase shifts of IOSs and the beamforming vectors of the BS. To address the varying dimensionality and the non-convexity of the optimization problem, we decompose it into two subproblems that optimize the IOSs’ phase shifts together with the BS beamforming vectors and the number and locations of IOSs, respectively, and devise successive convex approximation and Continuous Population-Based Incremental Learning-based algorithms to solve them alternately. Simulation results demonstrate that the proposed algorithms can obtain the optimal number and locations of IOSs, resulting in significantly enhanced energy efficiency of the outdoor BS compared to benchmark schemes.

The increasingly stringent requirement on quality-of-experience in 5G/B5G communication systems has led to the emerging neural speech enhancement techniques, which however have been developed in isolation from the existing expert-rule based models of speech pronunciation and distortion, such as the classic Linear Predictive Coding (LPC) speech model because it is difficult to integrate the models with auto-differentiable machine learning frameworks. In this paper, to improve the efficiency of neural speech enhancement, we introduce an LPC-based speech enhancement (LPCSE) architecture, which leverages the strong inductive biases in the LPC speech model in conjunction with the expressive power of neural networks. Differentiable end-to-end learning is achieved in LPCSE via two novel blocks: a block that utilizes the expert rules to reduce the computational overhead when integrating the LPC speech model into neural networks, and a block that ensures the stability of the model and avoids exploding gradients in end-to-end training by mapping the Linear prediction coefficients to the filter poles. The experimental results show that LPCSE successfully restores the formants of the speeches distorted by transmission loss, and outperforms two existing neural speech enhancement methods of comparable neural network sizes in terms of the Perceptual evaluation of speech quality (PESQ) and Short-Time Objective Intelligibility (STOI) on the LJ Speech corpus.

Streptococcus pneumoniae is the most common human respiratory pathogen, and β-lactam antibiotics have been employed to treat infections caused by S. pneumoniae for decades. β-lactam resistance is steadily increasing in pneumococci and is mainly associated with the alteration in penicillin-binding proteins (PBPs) that reduce binding affinity of antibiotics to PBPs. However, the high variability of PBPs in clinical isolates and their mosaic gene structure hamper the predication of resistance level according to the PBP gene sequences. In this study, we developed a systematic strategy for applying supervised machine learning to predict S. pneumoniae antimicrobial susceptibility to β-lactam antibiotics. We combined published PBP sequences with minimum inhibitory concentration (MIC) values as labelled data and the sequences from NCBI database without MIC values as unlabelled data to develop an approach, using only a fragment from pbp2x (750 bp) and a fragment from pbp2b (750 bp) to predicate the cefuroxime and amoxicillin resistance. We further validated the performance of the supervised learning model by constructing mutants containing the randomly selected pbps and testing more clinical strains isolated from Chinese hospital. In addition, we established the association between resistance phenotypes and serotypes and sequence type of S. pneumoniae using our approach, which facilitate the understanding of the worldwide epidemiology of S. pneumonia.

XML which is a new language for data representation is expected to become a universal format for data exchange on the Web. Generally speaking, we make some changes on XML documents as time goes by. Because temporal XML can store the successive versions of a document in an incremental fashion, it provides us a high effective method for version management. Of course, XML is more natural than RDB on representing the temporal information. So the topic of representing, querying and updating temporal information in XML has received some attention. In this paper a XML data model for tracking historical information in an XML document are proposed. After presenting an abstract model for bitemporal XML, two different ways of mapping this abstract representation into a (bitemporal) XML document are discussed. Finally all schemas are compared.

Information security and access control is an important part of CSCD applying. RBAC (role based access control) as an access control technology using for distributed system has been widely researched these years. Role hierarchies is one component of RBAC model, it can reduce the workload of permission assignment. Nowadays most research of RBAC has no relation with timing constraint. The main purpose of this article is to study the effect of temporal constraint acting on role hierarchy. The basis of the study is RBAC96 model, first we analyse the change of role state, propose 3 forms of temporal constraint and give the expression of temporal constraint. Then we present an analysis of the effects of temporal constraint on role enabling and role activation which has various implication on a role hierarchy, and at last we formally describe a timing restricted role hierarchies model

Time information management in workflow has been recognized as one of the most significant tasks in workflow management. The uncertainties in time and time-related constraints in workflow models should be taken into consideration. Based on introducing time constraints on elements in fuzzy-timing Petri nets, this paper propose a new workflow model named fuzzy temporal workflow nets (FTWF-nets). The calculation of temporal elements in FTWF-nets is given. Then time modeling and time possibility analysis of temporal phenomena in FTWF-nets are investigated. Finally an example is given to illustrate how to use these methods. Research results show that FTWF-nets can be used to model time information in those workflows, which have time uncertainty and have time constraints on resources and activities, and analyze the time possibility on some constraints.

In recent years, numerous Petri net based workflow models have been proposed. In the aspect of temporal modeling information, however, there remain some problems in these workflow models. In this paper, a new Petri Net modeling method is put forward, which systematically describes the temporal information and the relevant resources during a process of workflow activities. Finally, an example is provided to illustrate the workflow modeling process with TempWF-net.

Machine learning is the science of building predictors from data while accounting for the predictor's accuracy on future data. Many machine learning classifiers can make accurate predictions when the data is complete. In the presence of insufficient data, statistical methods can be applied to fill in a few missing items. But these methods rely only on the available data to calculate the missing values and perform poorly if the percentage of missing values exceeds a threshold. An alternative is to fill in the missing data by an automated knowledge discovery process via mining the WWW. This novel procedure is applied by first restoring missing information and next learning the parameters of the classifier from the restored data. Using a Bayesian network as a classifier, the parameters, i.e., the probabilities associated with the causal relationships in the network, are deduced using the knowledge mined from the WWW in conjunction with the data available on hand. The method, when tested with heart disease data sets from the UC Irvine Machine Learning Repository [UCI repository of machine learning databases], gave satisfactory results.

Temporal characters due to the alteration of salary policy and salary standards are discussed in this paper. According to these characters, the authors put forward a formalization inference model of temporal logic and present a discussion of the knowledge database driven by temporal knowledge, called the temporal-driven knowledge database.